TWO-PHOTON IMAGING OF GLIAL ION AND WATER HOMEOSTASIS
IN HEALTH AND DISEASE
Alexander S. Thrane & Vinita Rangroo Thrane
Supervisors:
Prof. Erlend A. Nagelhus Prof. Maiken Nedergaard
Letten Centre, Institute of Basic Medical Sciences & Centre for
Molecular Medicine Norway Faculty of Medicine
University of Oslo Norway
Center for Translational Neuromedicine Division of Glial Therapeutics
Department of Neurosurgery University of Rochester
USA
2013
© Alexander S. Thrane and Vinita Rangroo Thrane, 2013
Series of dissertations submitted to the Faculty of Medicine, University of Oslo No. 1573
ISBN 978-82-8264-531-7
All rights reserved. No part of this publication may be
reproduced or transmitted, in any form or by any means, without permission.
Cover: Inger Sandved Anfinsen.
Printed in Norway: AIT Oslo AS.
Produced in co-operation with Akademika publishing.
The thesis is produced by Akademika publishing merely in connection with the thesis defence. Kindly direct all inquiries regarding the thesis to the copyright holder or the unit which grants the doctorate.
J TABLE OF CONTENTS
1. ACKNOWLEDGMENTS……… 4
2. ABSTRACT………. 5
3. ABBREVIATIONS……….. 6
4. LIST OF PAPERS……….... 7
5. INTRODUCTION……….... 8
6. HYPOTHESES AND AIMS……… 37
7. METHODOLOGICAL CONSIDERATIONS………. 38
8. SUMMARY OF RESULTS………. 57
9. DISCUSSION AND FUTURE DIRECTION.………. 63
10. CONCLUSIONS………. 79
11. REFERENCES……… 80
12. PAPERS I-VI..……… 99
K 1. ACKNOWLEDGMENTS
‘In my own view, some advice about what should be known, about what technical education should be acquired, about the intense motivation needed to succeed, and about the carelessness and inclination toward bias that must be avoided is far more useful than all the rules and warnings of theoretical logic.’ Santiago Ramòn y Cajal in ‘Advice for a Young Investigator’ (1916).
This thesis is the result of our combined work at the Center for Translational Neuromedicine in Rochester, NY, and at the Letten Centre in Oslo.
We would like to express our deepest gratitude for all those who made this work possible. First, we would like to thank our families for their unwavering support, including Berit, Vijay, Kumud, Per, Bente, Rytik, Camilla and Hannibal. Second, we would like to thank our supervisors Erlend A. Nagelhus and Maiken Nedergaard for their invaluable advice, numerous discussions, and long hours of hard work helping shape this thesis. Third, we would like to thank all the colleagues and collaborators that helped us both in Oslo and Rochester, including Michael J. Chen, Nathan A.
Smith, Justin Chang, Maria L. Cotrina, Qiwu Xu (Jim), Ning Kang, Nanhong Lou, Takahiro Takano, Takumi Fujita, Fushun Wang, Lane Bekar, Salvador Pena, Arnulfo Torres, Yonghong Liao, Douglas Zeppenfeld, Benjamin Plogg, Jeffrey Iliff, Thiyagarajan Meenakshisundaram, Rashid Deane, Arthur Cooper, Abdelatiff Benraiss, Weiguo Peng, Gry F. Vindedal, Scott Kennedy, Phillip M. Rappold, Nadia N. Haj-Yasein, David Wang, Georg A. Gundersen, Rune Enger, Torgeir Holen, Øyvind Skare, Johannes P. Helm, Anna E. Thoren, Vidar Jensen, Øyvind Hvalby, Arne Klungland and Ole P. Ottersen.
This work was sponsored by PhD stipends from IMB (A.S.T.) and from MLS (V.R.T.), and also Fulbright scholarships for both of the candidates.
L 2. ABSTRACT
The vital role of glial cells in the nervous system has only been revealed relatively recently as these cells are electrically silent. Technical advances that allow direct visualization of glial activity have therefore been paramount to the recent glial renaissance. One such technique, two-photon laser scanning microscopy (2PLSM), forms the basis of our thesis. 2PLSM offers the necessary spatial and temporal resolution to study glial cells in intact living brain. Exploiting the strengths of 2PLSM, we chose to examine key roles of glial water and ion homeostasis in health and disease. Our first study (Paper I) delineated a novel aquaporin-4 (AQP4) mediated glial signaling pathway that is activated in response to hypo-osmotic brain edema, and which may contribute to the high mortality of this condition. In the second study (Paper II) we showed that AQP4 plays a crucial role in oxygen microdistribution during cortical spreading depression (CSD), a wave of tissue depolarization closely linked to migraine. In the next study (Paper III), we found that general anesthetics directly suppress astrocyte calcium signals, and this effect may explain some of the sedative effects of these drugs. In Paper IV we explored the paravascular circulation of cerebrospinal fluid (CSF) through brain, and found that this compartment served both as a highway for selective lipid transport and signal mediation between astrocytes. In Paper V we used 2PLSM to show a pathological activation of microglia in the advanced stages of hepatic encephalopathy (HE), which may be linked to the development of fatal brain edema in severe cases of this condition. Finally, in Paper VI we use a congenital disorder of ammonia handling to show that the immediate neurotoxicity of ammonia is mediated by a direct short- circuiting of astrocyte potassium buffering. This secondarily causes an impairment of inhibitory neurotransmission, and is not related to astrocyte swelling as was previously hypothesized. Blocking the downstream accumulation of chloride in neurons using Na+-K+-2Cl- cotransporter subtype-1 (NKCC1) inhibitor bumetanide, we were able to reverse ammonia neurotoxicity and thus develop a clinically relevant therapy. Taken together, using in vivo 2PLSM imaging we were able to elucidate important roles of glial cells in normal brain function and debilitating neurological disorders.
M 3. ABBREVIATIONS
ΔF/F0 - change in fluorescence normalized to time 0
1PLSM - one-photon laser scanning microscopy
2PLSM - two-photon laser scanning microscopy
AM - acetoxymethyl Aqp4 - aquaporin-4 gene AQP4 - aquaporin-4 protein ATP – adenosine triphosphate BAPTA - 1,2-bis(o-
aminophenoxy)ethane-N,N,N',N'- tetraacetic acid
BBB - blood-brain barrier CNS - central nervous system CSD - cortical spreading depression CSF - cerebrospinal fluid
DAPC - dystrophin-associated protein complex
EGABA - GABA reversal potential eGFP - enhanced green fluorescent protein
EPSP - excitatory post-synaptic potential
FITC - fluorescein isothiocyanate GABA - γ-aminobutyric acid GFAP - glial fibrillary acidic protein Gln - glutamine
Glu - glutamate
Glt1 - glutamate transporter 1 gene GLT1 - glutamate transporter 1 protein GS - glutamine synthetase
HE - hepatic encephalopathy ICP - intracranial pressure ISM - ion-sensitive microelectrode IP3 - inositol-1,4,5-triphosphate IP3R2 - IP3 receptor 2
KCC2 - K+-2Cl- cotransporter-2 KO – knock-out
[K+]o - extracellular potassium concentration
mRNA - messenger ribonucleic acid mGluR5 - metabotropic glutamate receptor 5
MSO - L-methionine sulfoximine NH4+ - ammonium
NH3 - ammonia
[NH4+]o - extracellular ammonium concentration
NAD+ - nicotinamide adenine dinucleotide
NADH - reduced NAD+ NKA - Na+-K+-ATPase
NKCC1 - Na+-K+-2Cl- cotransporter-1 NMR – nuclear magnetic resonance Otcspf-ash - ornithine transcarbamylase deficient mouse with sparse-fur and abnormal skin and hair
ROI - region of interest
RVD - regulatory volume decrease shRNA - short hairpin ribonucleic acid Slc12a2 – NKCC1 gene
Vm - membrane potential WT - wild-type
N 4. LIST OF PAPERS
PAPER I
Thrane AS, Rappold PM, Fujita T, Torres A, Bekar LK, Takano T, Peng W, Rangroo Thrane V, Enger R, Haj-Yasein NN, Skare Ø, Holen T, Klungland A, Ottersen OP, Nedergaard M and Nagelhus EA. Critical role of aquaporin-4 (AQP4) in astrocytic Ca2+ signaling events elicited by cerebral edema. Proc Natl Acad Sci.
2011;108 (2):846-51. (A.S.T., P.M.R. and T.F. are joint first authors, commented upon in Nat Rev Neurosci, February 2011).
PAPER II
Thrane AS, Takano T, Rangroo Thrane V, Wang F, Peng W, Ottersen OP, Nedergaard M and Nagelhus EA. In vivo NADH fluorescence imaging indicates effect of aquaporin-4 deletion on oxygen microdistribution in cortical spreading depression.
Resubmitted after review to J Cereb Blood Flow Metab. (A.S.T., T.T. and V.R.T are joint first authors)
PAPER III
Thrane AS, Rangroo Thrane V, Zeppenfeld D, Lou N, Xu Q, Nagelhus EA and Nedergaard M. General anesthesia selectively disrupts astrocyte calcium signaling in the awake mouse cortex. Proc Natl Acad Sci. 2012;109(46):18974-9. (A.S.T. and V.R.T are joint first authors)
PAPER IV
Rangroo Thrane V, Thrane AS, Plog B, Thiyagarajan M, GundersenGA, Iliff J, Deane R, Nagelhus EA, and Nedergaard M. Paravascular microcirculation facilitates rapid lipid transport and astrocyte signaling in the brain. Submitted to Nat Neurosci. (V.R.T and A.S.T. are joint first authors)
PAPER V
Rangroo Thrane V, Thrane AS, Chang J, Alleluia V, Nagelhus EA and Nedergaard M. Real-time analysis of microglial activation and motility in hepatic and
hyperammonemic encephalopathy. Neuroscience. 2012; 220:247-55. (V.R.T and A.S.T. are joint first authors)
PAPER VI
Rangroo Thrane V, Thrane AS, Wang F, Cotrina ML, Smith NA, Chen M, Kang N, Fujita T, Nagelhus EA and Nedergaard M. Ammonia compromises astrocyte potassium buffering and impairs neuronal inhibition without causing swelling in vivo.
Pending resubmission after review to Nat Med. (V.R.T and A.S.T. are joint first authors)
O 5. INTRODUCTION
“What is the function of glial cells in neural centers? The answer is still not known, and the problem is even more serious because it may remain unsolved for many years to come until physiologists find direct methods to attack it.” Santiago Ramon-y Cajal
These words by Cajal rang very true in both his day and for the century that followed (Ransom et al. 2003). Because scientists did not possess adequate tools to directly explore the functions of glial cells, they were quickly forgotten amidst the vast array of data accumulated on neurons, where electrophysiology provided concrete endpoints to study. Like many other discoveries in science, the realization of the importance of glial cells coincided with the invention and subsequent application of new techniques to the field. One such technique that has gained momentum since the 1990s is two-photon laser scanning microscopy (2PLSM) (Denk et al. 1990), and this is the primary method of investigation used in our thesis. Imaging in live animals has traditionally been dominated by techniques such as functional magnetic resonance imaging and positron emission tomography, which have relatively poor temporal and anatomical (spatial) resolution (Lauritzen 2005; Schafer et al. 2012). High-resolution analyses have, on the other hand, been restricted to light and electron microscopy in fixed tissue sections or slices. Thus, high temporal and anatomical resolution has long been incompatible with analyses of living animals (Misgeld and Kerschensteiner 2006). The introduction of 2PLSM changed this, and now allows anatomical structures as small as dendritic spines to be studied in living brains over days and even weeks (Holtmaat et al. 2009). 2PLSM studies in intact animals thus allow investigators to address important biological questions in clinically relevant manner, significantly shortening the distance from bench to bedside (Dirnagl 2006).
Glial cells are by far the most numerous cell-type in the brain, outnumbering neurons 10:1 (Ransom et al. 2003). However, until around 30 years ago glial cells were only thought to act as a form of glue providing structural support for neurons.
Glial involvement in normal brain function and their roles in neurological disorders is therefore a relatively new field of research. Additionally, studies that increase our understanding of glial cells are highly salient for the general field of neuroscience, as they can open up a completely new set of therapeutic targets for brain disorders (Nedergaard et al. 2010; Verkhratsky et al. 2012b). Glial cells have already been
P implicated in a number of diseases, including stroke, epilepsy, Alzheimer, Parkinson, spinal cord injury, traumatic brain injury, brain edema, macular edema and migraine (Fuhrmann et al. 2010; Manley et al. 2000; Reichenbach et al. 2007; Takano et al.
2007; Tian et al. 2005; Verkhratsky et al. 2012b). They have also been shown to have key roles in regulating physiological parameters critical for neuronal function, such as extracellular potassium ([K+]o) and other ion homeostasis, volume regulation, neurotransmitter recycling, cerebral blood flow and energy metabolism (Attwell et al.
2010; Caesar et al. 2008; Haj-Yasein et al. 2012; Kirischuk et al. 2012; Mathiesen et al. 2012; Nedergaard and Verkhratsky 2012; Paulsen et al. 1988b; Pellerin et al. 2007;
Takano et al. 2006; Wang et al. 2012a). More studies focusing on how glial cells can generate and influence central nervous system (CNS) disorders are therefore clearly needed. Our thesis aims to use the advantages of 2PLSM to help address this ‘glial knowledge gap’ with the hope of elucidating novel therapeutic approaches.
Two-photon imaging using femtosecond-pulsed lasers
Before introducing the neurobiological background for our thesis studies, we first want to review the central aspects of 2PLSM imaging. Starting with the basic principles of fluorescence imaging – light is a form of electromagnetic radiation that can be described as both a particle, called a photon, or a wave. The wavelength of light describes its energy content; the shorter the wavelength the higher the energy of individual photons. Fluorescence imaging relies on the interaction of light with a molecule that then absorbs the light energy, raising the molecule to a higher energy (excited) state. The molecule, called a fluorochrome, then loses some energy through mechanical vibration, before emitting the rest of its energy as a new photon and returning to its basal ‘unexcited’ state. The emitted photon is always of lower energy (red-shifted) compared to the exciting photon due to the energy lost in vibration (Christensen and Nedergaard 2011). In a biological setting, relevant fluorochromes are frequently bonded to macromolecules to attain desired chemical properties, such as lipid solubility, and are then called fluorophores (Paredes et al. 2008).
The resolution of traditional light microscopy has been significantly improved with the advent of the laser, which is an acronym for ‘light amplification by stimulated emission of radiation’ (Maiman 1960). In laser scanning microscopy, a collimated (parallel rays) coherent (in phase) beam of monochromatic (only 1 wavelength) photons is scanned across the specimen of interest, usually in a raster
HG pattern using galvanometric mirrors (Duemani Reddy et al. 2008). However, an early issue with laser scanning microscopy was that extensive light excitation occurred above and below the focal point of the microscope, introducing noise to the collected image. This issue was greatly improved by the introduction of confocal laser scanning microscopy, where a strategically placed pinhole effectively cuts out all light emitted from outside the focal point. The pinhole also gives confocal microscopy its other important feature; the ability to collect optical sections from the specimen. However, the major issue with confocal microscopy that has made it largely incompatible with in vivo studies is the large amount of light energy delivered to the tissue both in and around the focal point (Misgeld and Kerschensteiner 2006). Another aspect of this problem, is that due to light scattering at the wavelength required to excite biologically relevant fluorophores confocal microscopy is unable penetrate tissue much deeper than 50 μm (Christensen and Nedergaard 2011).
Figure 1. Optical principles of 1PLSM and 2PLSM. Left: Differences between pulsed 2PLSM and continuous wave 1PLSM imaging.
Right: Jablonski diagram of two-photon vs.
one-photon absorption and emission.
Although theoretically conceived almost a century ago, 2PLSM was only made possible in the 1990s through the development of powerful femtosecond pulsed lasers (Denk et al. 1990). 2PLSM is based on the principle that if a pulsed laser emits a high enough concentration of photons, 2 or more photons may excite the same molecule simultaneously. Denk et al. were the first to successfully apply this concept to laser scanning microscopy (Denk et al. 1990). Since multiple photons are used to excite the target, each individual photon can have a relatively low energy, i.e. long wavelength (Fig. 1). This has several advantages over conventional confocal
HH microscopy in that the longer wavelengths used, frequently in the infrared band (800- 1000 nm), achieve much deeper tissue penetration (500-1000 um) and exert much less photodamage (injury due to light) (Helmchen and Denk 2005). The latter issue becomes even more important when conducting fluorescent imaging, as the fluorophores used can generate free radicals when excessively stimulated by light causing phototoxicity (Bestvater et al. 2002; Christensen and Nedergaard 2011).
Finally, the transfer of light energy in 2PLSM is non-linear, and only in the focal point of the beam will a critical concentration of photons be achieved sufficient to excite (or damage) the tissue (Table 1). This means that 2PLSM has inherent optical sectioning and noise reduction properties, and thus does not require a pinhole to block out-of focus light as with confocal microscopy.
Table 1. Summary of the salient 2PLSM and 1PLSM parameters.
2PLSM 1PLSM
Laser wave Pulsed Continuous
Wavelength 800-1000 nm 400-600 nm
Tissue penetration 500-1000 um 20-50 um
Excitation Non-linear Linear
Phototoxicity Limited to focal point Entire cone of light Emission All photons useful Out-of-focus photons
generate noise XY resolution
Z resolution
½ λ (e.g. 450 nm) λ (e.g. 900 nm)
½ λ (e.g. 250 nm) λ (e.g. 500 nm)
Neuroglia – more than just glue
The major difference between glial cells and neurons are that glia lack electrical excitability and therefore cannot fire action potentials. The lack of an easily measurable form of activity in glia is also one of the major reasons for the previously neurocentric focus of brain research (Ransom et al. 2003). The term glia was first coined by Rudolf L. K. Virchow (1821-1902), who thought that glia were a connective tissue ‘glue’ whose function was to provide structural support to neurons.
Later, in 1851, Heinrich Müller made the first drawing of a glial cell in the retina, now known as the Müller cell. In the spinal cord, Jakob Henle was the first to describe star-like ‘stellate cells,’ which we now know as astrocytes. These findings were later elaborated upon by numerous eminent histopathologists, including Camillo Golgi (1843-1926) and Santiago Ramón y Cajal (1852-1934) (Verkhratsky and Butt 2007). We now know that glia are a heterogeneous group of cells that can generally
HI be divided into macro- and micro-glia on the basis of morphology. The latter cell type is of mesenchymal origin and migrates to the brain in embryology, differentiating to become the resident immune-cell of the CNS (Kettenmann et al. 2011). Macroglial cells are of ectodermal origin and the sub-types include oligodendrocytes in the CNS and Schwann cells in PNS, which both provide the myelin sheath to neuronal axons (Barres 2008). However, the most numerous macroglial cell type and the major focus of our thesis are astrocytes. These complex and heterogeneous cells perform many of the essential homeostatic functions that the more specialized neurons cannot perform (Fig. 2) (Matyash and Kettenmann 2010). Astrocytes are capable of complex signaling and recent discoveries have led many to suspect that these cells may play an active role in CNS signal transmission (Agulhon et al. 2008).
Figure 2. Cellular composition of the CNS and PNS. Left: Overview of the different cell types and their relative proportions in the human nervous system. Right: Early illustrations of cellular heterogeneity in the brain created using the silver impregnation method. Modified from Retzius, Biologische Untersuchungen, 1894.
Astrocytes – stars of the brain
Astrocytes were named because of their star-like appearance, owing primarily to their high expression of cytoskeletal glial-fibrillary acidic protein (GFAP) (Ransom et al.
2003). Astrocytes can be divided up into two main subgroups, stellate and radial (elongated) cells. Stellate astrocytes in the grey matter are called protoplasmic astrocytes, and in the white matter fibrous astrocytes (Matyash and Kettenmann 2010). Radial astrocytes are further divided into Müller cells (in the retina) and Bergmann glia (in the cerebellum) (Nimmerjahn et al. 2009; Reichenbach et al. 2007).
Radial astrocytes, apart from those in retina and cerebellum, are commonly present
HJ during development in the brain and are thought to be involved in neural migration (Regan 2007). Protoplasmic astrocytes are the cell-type seen most frequently in the cerebral cortex of adults, and our thesis focuses primarily on this sub-group (Oberheim et al. 2012). Astrocytes, although named after their star-like appearance in early histological preparations, are in fact more bush-like, and extend a multitude of fine processes that almost entirely fill a 3-dimesional domain (Fig. 3). This astrocyte domain usually encompasses thousands of synapses, multiple blood vessels and usually does not overlap with neighboring astrocyte domains except during pathological conditions such as epilepsy (Oberheim et al. 2008). Astrocyte processes that extend to synapses are termed peri-synaptic and those extending to blood vessels are called perivascular endfoot processes (Simard and Nedergaard 2004). Astrocytes are thus strategically located in a domain-structure that can integrate local synaptic activity with metabolic supply from the blood (Oberheim et al. 2008). Moreover, because the extracellular space is of very limited size in the CNS, astrocytes are also ideally constructed to influence both the content and volume of the extracellular space locally, regulating neurotransmitter and ion concentrations (Thorne and Nicholson 2006).
Figure 3. Astrocyte complexity distinguishes advanced nervous systems. Left: GFAP-stained protoplasmic astrocytes from mouse (inset) and human, illustrating the dramatic increase in the complexity of these cells through evolution. Right: Mouse (inset) and human astrocytes labeled with intracellular dye reveals a more ‘bush’ than ‘star’-like structure compared to GFAP staining. Scale bar 20 um. Modified with permission from Oberheim et al. 2009.
Finally, astrocytes are extensively coupled to each other via gap-junctions, thus forming a vast syncytium of cells. Recent research has indicated that this may
HK help astrocytes act as effective conduits for redistributing metabolites or ions between active and inactive synapses (Giaume et al. 2010; Rouach et al. 2008). All of these anatomic specializations would lead one to suspect that highly developed astrocytes might be a pre-requisite for more complex nervous systems. Interestingly, glial complexity increases much more steeply than neuronal complexity from lower to higher species, perhaps indicating that highly evolved astrocytes are necessary to cope with high synapse densities (Oberheim et al. 2006). Additionally, some disease states in brain and retina are characterized by a reduction in the glia/neuron ratio early in the disease, perhaps indicating that common neurological disorders are primary gliopathies (Verkhratsky et al. 2012b).
There has been a vast increase in our understanding of the subset of channels and receptors expressed by astrocytes in recent years. Notably, astrocytes express abundant channels and transporters for potassium (e.g. inward rectifying K+ channel Kir4.1, Na+-K+-ATPase, NKA), sodium, chloride, water (e.g. aquaporin-4, AQP4), metabolites (e.g. monocarboxylate transporters), amino acids (e.g. system N transporter 1, SN1) and neurotransmitters (e.g. glutamate transporter 1, GLT1) (Kirischuk et al. 2012; Lovatt et al. 2007; Paulsen et al. 1988b). These membrane proteins are principally involved in what we now believe to be key astrocyte functions, including [K+]o buffering, neurotransmitter clearance and recycling at the synapse (especially glutamate and γ-aminobutyric acid, GABA), cerebral blood flow regulation, water transport and feeding neurons with lactate via the glucose-lactate shuttle (Bak et al. 2006; Haj-Yasein et al. 2012; Iliff et al. 2012; Paulsen et al. 1988a;
Pellerin et al. 2007; Takano et al. 2006). Additionally, immunogold and in situ studies have indicated that astrocytes might be capable of directly influencing synaptic processing via vesicular or non-vesicular (e.g. connexin hemichannel) neurotransmitter release, or so-called ‘gliotransmission’ (Bezzi et al. 2004; Hamilton and Attwell 2010; Jourdain et al. 2007; Zhang et al. 2004). In conclusion, several lines of evidence indicate that astrocytes can actively modulate neuronal function and thus regulate information transmission in the CNS, perhaps coupling this to local metabolic factors (Attwell et al. 2010). These seminal observations have led to a revision of the classical neurocentric view of the synapse, and prompted the inclusion of astrocytes into a ‘tripartite synapse’ (Araque et al. 1999). Our thesis seeks to extend this work and use 2PLSM to address two related ways that astrocytes can influence synaptic function: the control of extracellular ion and volume homeostasis.
HL Astrocyte water homeostasis and AQP4
A tightly regulated water homeostasis is essential for normal brain function (Amiry- Moghaddam and Ottersen 2003). An excessive influx of water into the brain parenchyma will cause swelling, called brain edema. Because the brain is enclosed in a rigid cranium this swelling will raise intracranial pressure, which can then set up a vicious cycle where venous drainage from the brain is compressed and pressure rises further until the brain herniates through the base of the skull. Brain edema is therefore associated with significant mortality and morbidity, and current treatments are unfortunately limited to only partially effective and quite old approaches such as hyperosmotic infusions or corticosteroids (Grande and Romner 2012).
Water can be transported through biological membranes via several mechanisms. Initially, water transport was thought to occur exclusively through passive diffusion directly through the lipid bilayer. However, this assumption was dispelled through Nobel Prize winning work by Peter Agre and co-authors, who isolated an ancient family of membrane channels in erythrocytes that are highly permeable to water (Preston and Agre 1991; Preston et al. 1992). Subsequent work has shown that almost all animals and plants express a family of water channels, called aquaporins, and to date no less than 12 isoforms have been identified in mammals (Agre et al. 2004). More recently, it has also been shown that membrane transporters (e.g. EEAT) and ion channels can contribute to active and passive water flux (MacAulay and Zeuthen 2010), and these critical observations also need to be incorporated to generate a comprehensive model of transmembrane water transport (Fig. 4).
Figure 4. Multiple pathways for water transport across the lipid bilayer. Aquaporins and several types of ion channels can increase plasma membrane H2O permeability. Most cotransporters also move H2O with every cycle of substrate transport.
HM Aquaporins are a family of transmembrane proteins that selectively permeate water, and can thus increase water flux across cell membranes manifold (King et al.
2004). Water transport across aquaporins is mediated by osmotic gradients, and hence diffusion is passive and bi-directional. A sub-group of aquaporins also permeate other substances in addition to water, such as urea, ammonia or glycerol, but the physiological role of this metabolite permeability is incompletely understood (King et al. 2004). Aquaporins are expressed in a range of tissues including blood (erythrocytes), kidney (renal tubules), lens, retina and brain. The principal types of aquaporins expressed in brain are AQP1 (choroid plexus epithelial cells), and AQP4 (astrocytes and ependymocytes) (Agre et al. 2004). Some aquaporins appear capable of gating the degree of water permeability (e.g. AQP2 via vasopressin in kidney) (Moeller et al. 2011). However, it is somewhat controversial whether AQP4 is capable of gating. Several studies have indicated that phosphorylation of different sites on the AQP4 molecule can increase or decrease water permeability (e.g. in response to glutamate, erythropoietin, vasopressin and potassium) (Gunnarson et al.
2009; Gunnarson et al. 2008; Illarionova et al. 2010; Moeller et al. 2009; Niermann et al. 2001; Song and Gunnarson 2012). However, subsequent structural determinations have contradicted these observations (Mitsuma et al. 2010).
Figure 5. AQP4 expression in astrocytes is polarized to endfeet and peri-synaptic processes. Left:
Immunogold labeling for AQP4 (black dots) showing sub-pial endfeet and peri-synaptic enrichment.
Modified with permission from Nielsen et al. 1997. Right: Individual AQP4 channels permeate water.
4 AQP4 channels can assemble into tetramers, which have central pore that may permeate other molecules. Numerous AQP4 tetramers join together into orthogonal arrays of particles that almost completely cover astrocyte endfoot membranes.
HN AQP4, which is the main aquaporin explored in our thesis, is expressed to such a large extent by astrocytes that formations of AQP4 proteins clustered in so- called orthogonal arrays of particles can be seen by freeze fracture EM to cover almost the entire surface of astrocyte processes (Fig. 5) (Rash et al. 2004). AQP4 is thought to be permeable primarily to water, and no other solutes, although there is some debate whether the central pore formed between tetramers can act as a gas- channel (Fang et al. 2002; Wang et al. 2007; Wang and Tajkhorshid 2010).
Additionally, AQP4 is selectively enriched in distinct subcellular domains, including perivascular endfeet and peri-synaptic processes, indicating that AQP4 may have a role in dissipating osmotic gradients generated near the synapses to the vasculature (Fig. 6) (Amiry-Moghaddam et al. 2003a; Binder et al. 2006; Frydenlund et al. 2006;
Nagelhus et al. 1999; Nielsen et al. 1997). The peri-vascular localization of AQP4 in brain is controlled by the dystrophin-associated protein complex (DAPC), and disruption of this complex in a mouse model of myotonic dystrophy is associated with a loss of the perivascular enrichment (Amiry-Moghaddam et al. 2004; Frigeri et al.
2001; Vajda et al. 2004). However, a recent study that AQP4 anchoring mechanism may differ between macroglial cell types (Enger et al. 2012).
Figure 6. The subcellular localization of AQP4 is ensured by the DAPC-complex. Left: Illustration showing the expression pattern of AQP1 (blue) and AQP4 (red) in the human brain. Right: AQP4 is anchored to distinct membrane domains in astrocytes via the DAPC-complex. Adapted with permission from Amiry-Moghaddam and Ottersen 2003.
HO AQP4 has been implicated in the pathophysiology of a range of CNS disorders, including hypo-osmotic brain edema, stroke, epilepsy, hydrocephalus, migraine, and meningitis (Amiry-Moghaddam et al. 2003b; Binder et al. 2004; Binder et al. 2006; Manley et al. 2000; Padmawar et al. 2005; Papadopoulos et al. 2004;
Papadopoulos and Verkman 2005). Constitutive Aqp4 knock-out (KO) mice have for instance been shown to have reduced hypo-osmotic brain edema and reduced stroke infarct size, which is hypothesized to be related to decreased astrocyte swelling and secondary dysfunction (Manley et al. 2000). A glia-specific Aqp4 KO has also recently been created, and observations from these animals are strikingly similar to constitutive KO, indicating that most of the functional AQP4 is expressed in astrocytes and not e.g. endothelial cells (Haj-Yasein et al. 2011c). Although there is an expanding literature exploring AQP4 functions in the CNS, several critical questions remain unanswered, including: 1) Are all forms of astrocyte swelling AQP4-dependent, or is this phenotype only evident in cell culture or acute brain slices from young animals? 2) Does AQP4 play a role in astroglial signaling through sensitizing these cells to small volume changes near the synapse or blood vessels? 3) Does AQP4 mediate astrocyte volume changes at the level of the individual cell or is the role of AQP4 instead to facilitate more global water and cerebrospinal fluid (CSF) flux (Iliff et al. 2012)? 4) Does AQP4 only permeate water or can it also act as a gas channel to for instance O2? Because this is a rapidly evolving field, the current thesis has restricted itself to addressing only a subset of these questions.
Astrocyte regulation of [K+]o homeostasis in the brain
Rapid and efficient potassium homeostasis is vital in maintaining normal neuronal function, and is thought to be one of the primary functions of astrocytes (Kofuji and Newman 2004). Interestingly, the blood-brain barrier (BBB) is highly impermeable to potassium ions, ensuring that [K+]o in the CNS is regulated internally and is minimally affected by systemic fluctuations (Hansen et al. 1977). Astrocytes have multiple passive and active potassium transport mechanisms, and similar to most mammalian cells the potassium gradient is largely responsible for their negative (more so than neurons, ~85 mV) membrane potential (Kofuji and Newman 2004).
Moreover, astroglia including Müller cells likely have a much greater permeability to and capacity for buffering [K+]o than neurons, although this permeability may be
HP restricted to functional micro-domains (Newman et al. 1984; Walz 2000). Astrocyte potassium buffering is thought to be divided into two main mechanisms: local potassium uptake and spatial buffering (Fig. 7) (Kimelberg and Nedergaard 2010;
Kofuji and Newman 2004).
Figure 7. Two proposed mechanisms for astrocyte potassium homeostasis. Top:
Potassium spatial buffering involves astrocyte uptake of increased potassium from the synaptic cleft down its electrochemical gradient to the vasculature. This mechanism is thought to involve potassium channel Kir4.1 and potentially AQP4. Bottom: Local potassium accumulation coupled with sodium pumping and sodium/chloride cotransport to remove excess K+ released from neuronal activity. This mechanism is thought to involve NKCC1 / KCC2 and the NKA. Modified with permission from Kimelberg and Nedergaard 2010.
Ion pumps, cotransporters or potassium channels can all mediate local potassium uptake. This mechanism is thought to be temporary and the influxed potassium ions are eventually released back into the extracellular space, restoring the ionic composition of the various cellular compartments. Cotransporters and ion pumps usually couple K+ influx to anion (e.g. Cl-) influx or cation (e.g. Na+) efflux, to maintain electroneutrality. The NKA pump and (Na+)-K+-Cl- cotransporters are thought to be the main proteins responsible for short-term buffering of increases in [K+]o following neuronal activity (Andersson et al. 2004; Kofuji and Newman 2004;
Miyakawa-Naito et al. 2003; Olson et al. 1986; Xiong and Stringer 2000). The NKA exports 3Na+ and imports 2K+ causing it to be mildly electrogenic (hyperpolarizing the membrane potential, Vm) and is sensitive to inhibition by several drugs, such as the plant-toxins ouabain and digitalis. Previous studies have indicated that the astroglial isoform of NKA is more effective in buffering [K+]o than the neuronal isoform (Grisar et al. 1983; Reichenbach et al. 1987). (Na+)-K+-Cl- cotransporters move all relevant ions in an electroneutral manner, and are highly expressed in secretory epithelia and ependymal cells, and to a lesser extent in both neurons and astrocytes (Delpire et al. 1999; Rivera et al. 2005). There are several important types
IG of (Na+)-K+-2Cl- cotransporters, including Na+-K+-Cl- cotransporter-1 (NKCC1) expressed at a high level in early development when it imports Cl-, and K+-2Cl- cotransporter-2 (KCC2) expressed later in development when it exports Cl-. The ontogenic expression patterns of these two cotransporters are thought to represent a developmental switch in the action of GABA, from depolarizing in early development to hyperpolarizing in adulthood (Rivera et al. 2005; Rivera et al. 1999). NKCC1 but not KCC2 is highly sensitive to low doses of the clinically used diuretic bumetanide, whilst both transporters are inhibited by the diuretic furosemide (Kimelberg 1987;
Kofuji and Newman 2004).
The second mechanism for astroglial for potassium clearance is spatial buffering (Chen and Nicholson 2000; Xiong and Stringer 2000). This mechanism entails astrocytes redistributing local increases in [K+]o from areas of high neuronal activity through the gap junction-coupled astrocyte syncytium to areas of less activity with lower [K+]o. The driving force for the potassium current is the difference between the local potassium equilibrium potential (EK) and the glial syncytium membrane potential (Vm) (Chever et al. 2010; Djukic et al. 2007; Haj-Yasein et al.
2011a; Heuser et al. 2012; Karowski and Proenza 1977; Nagelhus et al. 2004;
Newman et al. 1984). Potassium siphoning is a variant of spatial buffering that involves redistribution of increased [K+]o to specific subcellular compartments due to polarized potassium channel expression, such as in retinal Müller cells (Newman 1985; Newman et al. 1984). One of the main advantages of this mechanism is that it allows the redistribution of potassium ions with little increases in potassium inside individual astrocytes. However, though this model provides an astrocyte-specific mechanism for buffering [K+]o, there are many faults with the theory including the lack of a sufficiently strong electrochemical gradient to push a potassium current between astrocytes (Meeks and Mennerick 2007). In the brain the potassium channel subtype Kir4.1 is highly enriched in astrocytes, where it has a polarized expression similar to AQP4, principally being present in endfeet and peri-synaptic membranes. It therefore seems likely that there is some functional coupling between potassium and water homeostasis in these specialized astrocyte membrane domains (Amiry- Moghaddam et al. 2003b; Chever et al. 2010; Djukic et al. 2007; Nagelhus et al. 1999;
Strohschein et al. 2011). Any spillover or siphoning of excess potassium beyond the local uptake at the individual synapse is thought to be a hallmark of pathology or an artifact of experimental preparations (e.g. epilepsy or spreading depression) (Kofuji
IH and Newman 2004). Potassium homeostasis in the brain is a controversial topic and further studies are clearly warranted to help ascertain the relative importance of the various mechanisms involved in [K+]o buffering.
Cortical spreading depression (CSD) and migraine
CSD is a self-propagating wave of depolarization and neurodepression that spreads through cortical grey matter at a slow rate of 2-5 mm min-1 (Lauritzen 2001; Leao 1947). Waves of CSD can be evoked in experimental animals by intense neuronal stimulation, local application of potassium or tissue injury (Takano and Nedergaard 2009).
Figure 8. Model of migraine pathophysiology. The illustration shows a human cortex with a wave of CSD spreading across it. The CSD wave itself (yellow) is accompanied by neurodepression (decreased electrocorticogram amplitude), depolarization (increased [K+]o), hyperemia, hypoxia and will typically manifest as visual phenomena called an aura. This period is followed by prolonged oligemia, slow restoration of neuronal activity and activation of trigeminal nociceptive afferents causing the characteristic headache (green). Modified with permission from Takano et al. 2009.
CSD has been closely linked to the pathophysiology of migraine headaches, where sufferers often experience a phenomenon called aura that manifests as sensory phenomena (e.g. scintillation-scotoma). These phenomena show a slow somato- or retino-topic spread across the cortex that is of an equivalent velocity to CSD
II (Lauritzen et al. 2011). As the ‘aura wave’ subsides, the patient typically begins to experience throbbing headaches, which is thought to be related to secondary meningeal vasodilation with activation of trigeminal nociceptive afferents (Fig. 8) (Takano and Nedergaard 2009). However, the characteristic wave of depressed neuronal (electrocorticogram) activity seen during CSD is present in a very narrow zone of cortex, and thus cannot be detected non-invasively with external electroencephalogram electrodes in humans (Lauritzen et al. 2011). When functional imaging is used to measure cerebral blood flow during migraine attacks investigators again observe a brief and variable hyperemia (more prominent in anesthetized animals) followed by a characteristically slow wave of cortical oligemia (Lauritzen 1994; Tfelt-Hansen and Koehler 2011). Finally, using more invasive recordings investigators have shown that human cortex is capable of generating and sustaining CSD waves following injury or experimental stimulation (Lauritzen 1994). Taken together, CSD is therefore likely to be a fundamental property of all densely packed nervous tissue, both animal and human.
CSD is characterized by an increase in [K+]o of more than 50 mM, and re- establishing the transmembrane potassium gradient thus presents a unique metabolic challenge for the brain parenchyma (Takano et al. 2007). Additionally, other electrolyte gradients are critically altered, with [Ca2+]o decreasing from 1.2 to 0.1 mM, [Cl-]o decreasing from 120 to 70 mM and [Na+]o decreasing from 150 to 60 mM (Lauritzen et al. 2011). CSD waves are usually self-terminating and do not cause permanent brain injury (Takano et al. 2007). However, in the context of severe cortical injury (e.g. traumatic brain injury or stroke), so-called peri-infarct or anoxic depolarizations can be observed that persist and increase the area of brain that is permanently damaged (Lauritzen et al. 2011). Interestingly, when fluorocitrate is used to selectively injure astroglial by inhibiting the Krebs cycle this also causes CSD to induce irreversible neuronal dysfunction (Largo et al. 1997; Paulsen et al. 1987). The predilection for migraine attacks to originate in the striate (V1) cortex in humans may therefore relate directly to the relatively lower astrocyte-to-neuron ratio in this area (Lauritzen et al. 2011).
To study this metabolic strain we chose to employ reduced nicotinamide adenine dinucleotide (NADH) imaging, which relies on the natural auto-fluorescence of NADH but not NAD+ (Fig. 9). NADH is critical co-enzyme in cellular metabolism, which is generated from NAD+ (reduction) during glycolysis, and to a larger extent
IJ from Krebs cycle, and is used by the mitochondrial respiratory chain (oxidation).
When used in conjunction with 2PLSM, NADH imaging can thus be used to generate detailed maps of local tissue red-ox state, where increased NADH signal intensity is a sensitive proxy of tissue hypoxia (Kasischke et al. 2011). Interestingly, NADH fluorescence in neuropil is highest in astrocytes, almost precisely outlining the entire cells (including processes), whilst showing little overlap with mitochondrial staining (Kasischke et al. 2004). NADH imaging during CSD or hypoxemia reveals sharply defined cylindrical perivascular regions of high oxygen tension (low NADH signal) and irregularly shaped interspersed watershed regions of low oxygen tension (high NADH signal) (Kasischke et al. 2011; Takano et al. 2007). The steep sigmoidal NADH-pO2 curve with a p50 for O2 of 3.4 mmHg explains the sharp border between high and low NADH signal intensity in perivascular and watershed regions, respectively (Kasischke et al. 2011).
Figure 9. 2PLSM imaging of tissue red-ox state using NADH fluorescence. Left: NADH, but not NAD+, is fluorescent and serves as a sensitive proxy of tissue hypoxia in the brain. Right: NADH imaging reveals increased NADH indicative of tissue hypoxia in geometrically shaped microwatershed regions of mouse cortex during brief and mild hypoxemia. Arteries and veins are pseudo-colored in red and blue, respectively. Modified with permission from Kasischke et al. 2011.
Using in vivo 2PLSM NADH imaging Takano et al. in 2007 made a set of key observations regarding the micro-distribution of oxygen in CSD. The authors observed that in watershed areas tissue hypoxia causes and coincides with CSD, and critical hypoxia in these regions causes swelling along with dendritic spine distortion in neurons, but not astrocytes (Takano et al. 2007). The extent of astrocyte swelling is still somewhat more debated, but the much clearer swelling in neurons shown in this and other studies at least indicates that astrocyte volume regulation is much tighter (from unpublished observations in situ) (Risher et al. 2012) (Takano et al. 2007; Zhou
IK et al. 2010). More importantly, the NADH observations made by Takano et al.
indicate that the self-propagating nature of the CSD wave might stem from a primary energy failure (i.e. NKA failure). This energy failure seems most likely to stem from excess energy consumption than from inadequate supply, as blood flow actually increases during the early phase of CSD. However, what sets up the CSD wave in the first place is a bit more uncertain. The wave itself is preceded by propagating field oscillations in glial and neuronal Vm, indicating that CSD might be initiated by neuronal hyperexciteability similar to seizures (Larrosa et al. 2006). Additionally, genetic analysis of patients with familial hemiplegic migraine has revealed mutations in genes encoding calcium channels (e.g. Cav2.1), the astrocytic Na+-K+-ATPase and sodium channels (e.g. SCN1) linked to maintaining Vm (Leo et al. 2011; Takano and Nedergaard 2009). In conclusion, CSD presents a unique metabolic challenge for astrocytes and neurons alike, and the self-sustaining slow propagation of the CSD wave appears to be caused by a vicious cycle of excess oxygen consumption, watershed hypoxia and a secondary failure of potassium homeostasis (Takano et al.
2007).
Calcium signaling in astrocytes
The advent of calcium imaging using chemical calcium indicators (e.g. Fluo-3) was one of the major breakthroughs that alerted investigators to the active role of glia in brain information processing. It was known through histological studies since the 1980s that astrocytes express a subset of glutamate receptors (Bowman and Kimelberg 1984; Kettenmann et al. 1984). However, it was not until 1990 that Cornell-Bell et al. observed that cultured astrocytes are able to respond to locally applied glutamate by generating wide-spreading waves of elevated intracellular calcium (Cornell-Bell et al. 1990). This observation was critical to subsequent glial research because it showed that astrocytes have the ability to directly sense synaptic activity. Subsequently, co-cultures of astrocytes and neurons were used to show that astrocytes are not just able to respond, but also to alter neuronal signaling through intracellular calcium signals (Nedergaard 1994; Parpura et al. 1994). Finally, these observations of bidirectional astrocyte communication with neurons were validated in more intact preparations (acute hippocampal slices and retinal preparations) (Kang et al. 1998; Newman and Zahs 1998). More recent work has also indicated that in response to the activation of glutamatergic and purinergic receptors or alterations of
IL local ion concentrations astrocytes may be able to release neurotransmitters into the synaptic cleft, either via vesicular or non-vesicular mechanisms (e.g. connexin hemichannels) (Hamilton and Attwell 2010; Kang et al. 1998; Simard et al. 2003;
Torres et al. 2012; Wang et al. 2012a; Wang et al. 2012b). This extensive astroglial signaling at the level of the individual synapse led to the development of the ‘tripartite synapse’ model, which incorporates an astrocytic process into the basic unit of information processing (Araque et al. 1999).
Figure 10. Illustration of the molecular mechanisms governing intracellular calcium in astrocytes and neurons. Abbreviations: CBP: Ca2+-binding protein; ER: endoplasmic reticulum; GLT: glutamate transporter; GPCR: G-protein coupled receptor; InsP3R: inositol-1,4,5-trisphosphate receptor; LGCC:
ligand-gated Ca2+ channel; NCX: sodium–calcium exchanger; PMCA: plasmalemmal Ca2+ ATP-ase;
RyR: ryanodine receptor; SERCA: sarco(endo)plasmic reticulum Ca2+ ATP-ase; SOC: store-operated Ca2+ channel; VGCC: voltage-gated Ca2+ channel; ‘Maxi’ channel: high-permeability plasmalemmal channels (such as connexins or pannexins or P2X7 receptors or volume-regulated anion channels) that can act as a pathway for non-exocytotic gliotransmitter release. Modified with permission from Nedergaard et al. 2010.
Intracellular transients of free ionized calcium (Ca2+) are an old form of cellular signaling that is present even in bacteria, and likely co-evolved with the use of adenosine triphosphate (ATP) as the primary intracellular energy source (Dominguez 2004). Intracellular calcium is tightly regulated in all mammalian cell types, and a rapid increase in this ion is used to among other things to mediate programmed cell-death (apoptosis) (Parpura and Verkhratsky 2012). Astrocytic calcium signals appear to derive primarily from endoplasmic reticulum (ER), where
IM the concentration of calcium is ~1000x higher than in the cytoplasm. Upon activation of metabotropic Gq-coupled G-protein coupled receptors, phospholipase C cleaves phosphatidylinositol-4,5-bisphosphate in plasmalemma, releasing inositol-1,4,5- trisphosphate (IP3) to the cytosol and diacylglycerol in the membrane. IP3-dependent calcium release from the ER via the IP3 receptor type 2 (IP3R2) is thought to be the major pathway of calcium signaling in astrocytes; as opposed to neurons where voltage gated calcium channels mediate vesicular neurotransmitter release (Fig. 10) (Nedergaard et al. 2010). However, it is possible that smaller IP3-independent calcium signals in the small processes trigger IP3 signaling in so-called calcium-induced calcium release (Arcuino et al. 2002). IP3-independent mechanisms for calcium influx in astrocytes include plasmalemmal calcium channels (e.g. Cav1) and/or transporters (e.g. Na+-Ca2+-exchanger) (Nedergaard et al. 2010).
Intracellular calcium transients are known to propagate as waves throughout a network of astrocytes joined by gap junctions (composed of two connexons), which is termed the astrocytic syncytium (Rouach et al. 2008). There are a number of mechanisms thought to be involved in calcium wave propagation, including the diffusion of Ca2+ or IP3 through gap junctions and paracrine ATP release activating neighboring metabotropic P2Y receptors (Verkhratsky et al. 2012a). Astrocytic calcium waves as observed in vitro and in situ propagate relatively slowly compared to axonal signals (4-20 μm s-1 vs. 10-100 m s-1) (Hartline and Colman 2007). For this reason, some authors would argue that astrocytic calcium waves may be too slow or could generate noise if they were to modulate individual synaptic events (Agulhon et al. 2008). Conversely, this temporal difference may indicate that astrocyte calcium transients are responsible for the integration of neuronal activity over a larger temporal (seconds) and spatial (thousands of synapses) scale, which is more relevant for neurovascular coupling (Iadecola and Nedergaard 2007).
Through the application of 2PLSM to calcium imaging, investigators have recently been able to explore astrocyte calcium signals in intact living brains. In one of the seminal in vivo studies Wang et al. showed that physiological sensory stimulation (in the form of air-puffing whiskers) is able to elicit a coordinated calcium response that followed neuronal firing in the stimulated area of somatosensory cortex (Wang et al. 2006). The magnitude of this sensory-evoked astrocytic calcium response was directly correlated with the neuronal response and depended on
IN activation of Gq-coupled metabotropic glutamate receptor 5 (mGluR5). Subsequent in vivo studies have elaborated on this observation, and shown that other forms of sensory stimulation (e.g. hindlimb electroshock), as well as local agonist application or stimulated neuromodulator release (e.g. acetylcholine and norepinephrine) can also elicit calcium activity in cortical astrocytes (Bekar et al. 2008; Navarrete et al. 2012;
Takata et al. 2011; Winship et al. 2007). Astrocytes thus appear able to respond to several types of physiologically relevant signals in vivo.
Figure 11. Astrocytic regulation of blood flow via calcium signals. Left: Illustration of the hypothetical role of astrocytes in connecting neuronal demand with blood flow supply via calcium signaling. Adapted with permission from Iadecola and Nedergaard, 2007. Right:
Serial 2PLSM images taken at 3 s intervals of a cerebral arteriole showing functional hyperemia and associated calcium activity (pseudocolored rhod-2 intensity) in the awake mouse cortex (unpublished).
Since in vitro and in situ work showed astrocyte-neuronal signaling is inherently bidirectional, the next question would be what neuronal changes do astrocyte calcium signals elicit in vivo. Unfortunately, on this critical issue the literature is quite divided. The previously discussed ‘tripartite synapse’ model states that astrocyte calcium activity triggers gliotransmitter release, which may in turn modulate neuronal activity (Araque et al. 1999). However, data that directly questions some of the key observations underpinning this model has emerged more recently (Agulhon et al. 2010; Agulhon et al. 2008; Nedergaard and Verkhratsky 2012; Sun et al. 2013). Another critical function for astroglial calcium transients is mediating neurovascular coupling (functional hyperemia) (Fig. 11). Stimulating astrocyte calcium transients or experimentally increasing intracellular calcium using uncaging triggers brisk vascular responses. However, both the type of response (vasodilation or constriction) and the mechanisms implicated are currently under intense debate, as
IO different experimental paradigms have generated divergent results (e.g. pre- constricted blood vessels in brain slice vs. in vivo) (Attwell et al. 2010; Iadecola and Nedergaard 2007; Lauritzen et al. 2012; Takano et al. 2006).
There are several important questions regarding astrocyte calcium signaling that remain unanswered. First, can we trust the data from in situ studies of young animals or are hippocampal slices not as good a model study astrocytes as it has been for neurons (Sun et al. 2013)? Second, in vivo imaging has a much higher signal-to- noise ratio for calcium recordings and is often acquired at a lower temporal resolution (e.g. 1-10 Hz frame rate). This raises the question of whether we are in fact detecting all the relevant calcium signals or looking in the right places (e.g. soma v.s.
processes). Two in situ studies have recently shown that short lasting low amplitude calcium transients in the processes occur even in response to individual synaptic events (Di Castro et al. 2011; Panatier et al. 2011). More recently, high-speed 2PLSM set-ups developed primarily for neuronal calcium imaging are breaching this barrier, allowing frame rates of up to 1000 Hz (Chen et al. 2011). Third, does neuronal activity generate multiple astrocytic calcium waves, and if so do these have different kinetics and functions (Torres et al. 2012)? Fourth, are we imaging deep enough in the cortex? Whisker activation principally stimulates layer IV, and we may be simply observing an axially propagated response in the more superficial layers. Finally, one major problem with the in vivo literature to date is that, with the exception of two studies, it has been carried out using anaesthetized animals, where neuronal activity is known to be significantly depressed (Greenberg et al. 2008). These two preliminary studies of calcium signaling in awake animals revealed profoundly different spontaneous and evoked activity patterns. These included synchronized ‘bursts’
(instead of waves) across hundreds of astrocytes and a much higher level of spontaneous activity (Dombeck et al. 2007; Nimmerjahn et al. 2009). Exploring calcium signaling in astrocytes in awake animals would no doubt give a more physiological view of how astrocytes signal and perhaps provide some more definitive answers to whether or how astrocytes can sense and modulate synaptic transmission.
Cerebrospinal fluid microcirculation in the brain
In most tissues of the body, a parallel lymphatic circulation accompanies blood vessels and supports normal hemodynamic function by resorbing interstitial fluid,
IP carrying immune cells and transporting lipids (Schmid-Schonbein 1990). The brain, however, is one of the few organs to be entirely devoid of an anatomically distinct lymphatic system. Through the selective barrier properties of the BBB, the composition of the CNS microenvironment is regulated very separately from the rest of the body (Abbott 2004). It has long been thought that CSF might serve the role of peripheral lymph in the CNS. Macroscopically, CSF circulates rapidly from the choroid plexi in the ventricles where it is produced and diffuses through the brain parenchyma by convective bulk flow, before being resorbed by the arachnoid granulations into venous sinuses (Figure 12) (Abbott 2004; Aukland and Reed 1993;
Sakka et al. 2011). However, the extracellular space in the brain is extremely narrow and very tortuous, meaning that any bulk flow through this anatomical space should be much slower than the interstitial fluid diffusion rates we observe (0.1-0.3 µL min-1 g-1) (Thorne and Nicholson 2006).
Figure 12: Paravascular CSF circulation in neuropil. Left: Macroscopic CSF circulation, illustrating para-arterial (Art., red) and para-venous (Ven., purple) PVS (blue) ensheathed by astrocyte endfoot (e.f.) processes. Paravascular CSF flow mediates convective bulk flow of interstitial fluid through the narrow and tortuous extracellular space. Right: Microscopic view of paravascular circulation of interstitial fluid (ISF) and lipids through putative transporters in glial endfeet.
These observations have led investigators to suggest that ‘highways’ for more rapid CSF flow through the parenchyma must exist, and the two most likely candidates are the para-axonal and paravascular pathways. The former is more highly developed in species with bigger axons, but diffusion-weighted MRI has also revealed that water flux along axons is a major pathway in the human brain (Abbott 2004;
JG Mori et al. 2009). Several groups have illustrated the latter pathway since the 1800s by injecting various tracer molecules (e.g. horseradish peroxidase) injected into the brain parenchyma, ventricles or subarachnoid CSF and performing serial histological sectioning. This method causes the tracer to strikingly outline a paravascular space (PVS) ensheathing the cerebral vasculature, although the choice of tracer determines how completely the vasculature will be outlined (Abbott 2004; Cserr et al. 1981;
Cserr et al. 1986; Rennels et al. 1990; Szentistvanyi et al. 1984). Interestingly, a recent 3D EM reconstruction has revealed that astrocyte endfeet provide a complete ensheathment of cerebral blood vessels, ensuring the patency of this paravascular compartment (Mathiisen et al. 2010).
The paravascular CSF pathway was recently revisited by Iliff et al., who used 2PLSM to demonstrate that a range of water-soluble tracers and deuterated water (2H2O) itself pass rapidly along the separate anatomical space between astrocyte endfeet and the blood vessel wall in vivo (Iliff et al. 2012). The PVS is largest near the pial surface, where it is called the Virchow-Robin space, and becomes progressively narrower before almost disappearing at the capillary level (Abbott 2004).
Interestingly, Iliff et al. also showed that water flow through this pathway was 70%
slowed by deletion of Aqp4, indicating that one of the major roles of this protein could be facilitating CSF microcirculation (Fig. 13). It therefore appears that the brain has a distinct pathway for the microcirculation of water and water-soluble compounds. However, another major function of peripheral lymph is the transport of lipids from the gut to the liver and peripheral organs. Despite grey and white matter being composed of 38 and 58 % lipid, respectively (O'Brien and Sampson 1965a;
O'Brien and Sampson 1965b), the BBB is largely impermeable to peripheral lipids and lipoproteins (Leoni et al. 2010). The essential lipids needed for myelin or synaptic membranes therefore need to be produced within the CNS (Mauch et al. 2001).
Interestingly, astrocytes have been shown to orchestrate this vital process, by producing cholesterol and high-density lipoprotein-like molecules (Fagan and Holtzman 2000). The addition of fatty acid residues (palmitoylation) has important functions during neurodevelopment and synaptic plasticity (Fukata and Fukata 2010).
Astrocyte lipid clearance to the blood is also likely to be important, and for instance cholesterol is cleared by conversion to 24S-OH-cholesterol (cerebrosterol), making it BBB permeable (Bjorkhem 2007). CSF is rich in astrocyte-produced lipoproteins, and mutations in genes related to this process (e.g. ApoE) have been linked to
JH neurodegenerative conditions such as Alzheimer disease (Leoni and Caccia 2011;
Leoni et al. 2010; Riddell et al. 2008). Taken together, observations regarding lipid metabolism in the CNS indicate the existence of astrocyte-neuronal lipid (or cholesterol) shuttle, where astrocytes feed neurons with the necessary lipid for synaptic membranes (Nieweg et al. 2009). An important question that is unanswered in current literature is therefore whether the PVS serves as a conduit for rapid lipid as well as water microcirculation through the brain parenchyma.
Figure 13. Immunofluorescence micrograph of AQP4 and GFAP expression in the mouse cortex.
Modified from Simard et al. 2003.
Microglia - the double-edged sword of CNS immunity
Microglia are a cell type that has come into focus relatively recently. They are the innate immune cells of the CNS and are of mesenchymal origin. Microglia are derived from peripheral macrophages and migrate to the brain during early development (Chan et al. 2007; Verkhratsky and Butt 2007). The function of microglia has been described as a ‘double-edged sword,’ as they have the capability of being both protective and harmful in health and disease. Under physiological conditions, microglia exist in the ‘resting state,’ which is morphologically characterized by the cells extending multiple thin processes from a small central cell body. However, in vivo 2PLSM recently showed that ‘resting’ microglial processes are in fact constantly moving, screening their territories, seeking out any signals that may indicate injury or inflammation (Davalos et al. 2005; Nimmerjahn et al. 2005).
Like astrocytes, microglia show very little overlap between cells and seem to survey their own domain with their processes (Kettenmann et al. 2011).
JI
Figure 14. The changing states of microglia. Illustration showing resting microglia (left), microglial activation and BBB opening after a small injury (middle), and severe microglial activation (right) associated with amoeboid morphological appearance and phagocytic behavior in the context of a large injury. Note the surveillance capability of fine processes in resting microglia, thought to constantly scan for foreign antigens.
In response to brain injury, infection or inflammation, microglia are known to change their morphology by transforming through an ‘alerted state’ to an ‘amoeboid state’ (Fig. 14). This morphological transition involves retraction of processes, which become fewer but thicker, increase in cell body size and production various cytokines along with other signals (Block et al. 2007; Davalos et al. 2005; Hanisch and Kettenmann 2007; Verkhratsky and Butt 2007). When activated, their morphological change aids phagocytosis and allows the removal of unwanted pathogens or dead host cells. The exact mechanism of how microglial activation is initially triggered is not fully understood, but we do know that microglia possess a number of neurotransmitter (including glutamate, GABA, adenosine and purine) and immune (including complement, cytokine and chemokine) receptors. Many of these receptors are involved in activating microglia during brain injury, but ATP in particular is thought to play a central role (Davalos et al. 2005; Kettenmann et al. 2011; Ohsawa et al.
2007). Brain injury is thought to result in the release of vast amounts of ATP, which microglia respond swiftly to through activation of mainly P2X7 receptors (Burnstock and Verkhratsky 2010; Kettenmann et al. 2011; Verkhratsky and Butt 2007).
Activation of these receptors promotes extension of microglial processes towards the perceived site of injury and the release of various cytokines and pro-inflammatory proteins. In addition to releasing pro-inflammatory markers (such as interleukins, tumor necrosis factor-α and transforming growth factor-β), microglia are thought to aid neurons in regeneration post-injury (Kettenmann et al. 2011). For example, they release a number of growth factors, including brain-derived neurotrophic factor and
